Detecting English (or other languages) text

in #python5 years ago

Determining if a message is English (or any language for that matter) can be accomplished by simply loading a dictionary, calculating percentage of words that are in the dictionary, calculating percentage of letters in the message, and checking that these percentages exceed a threshold.

First you need to load your dictionary file. I'm using /usr/share/dict/american-english as my dictionary, but you can pass in your own in the file argument.

import string
ENGLISH_WORDS = {}
def load_language_file():
    with open(file, 'r') as dictionary:
        word_list = dictionary.read()
    for word in word_list.split("\n"):
        ENGLISH_WORDS[word.upper()] = True

load_language_file("/usr/share/dict/american-english")

Take your message and view it as a list of words

words_in_msg = msg.split()

Count letters/words

    counts = {'words':0, 'letters':0}
    for word in words_in_msg:
        if word.upper().strip(string.punctuation) in ENGLISH_WORDS:
            counts['words']+=1
    for char in list(msg):
        if char in string.ascii_letters:
            counts['letters']+=1

Calculate word and letter percentage

calc_word_percentage = counts['words']/len(words_in_msg)*100
calc_letter_percentage = counts['letters']/len(msg)*100

Is the message valid? Compare the percentages to thresholds. In this case the message must be at least 40% words and 60% letters

word_percentage=40
letter_percentage=60
isValidMsg = calc_word_percentage > word_percentage \
        and calc_letter_percentage > letter_percentage

Full Source with text_matches_language function

import string
ENGLISH_WORDS = {}
def load_language_file(file):
    with open(file, 'r') as dictionary:
        word_list = dictionary.read()
    for word in word_list.split("\n"):
        ENGLISH_WORDS[word.upper()] = True
def text_matches_language(msg, word_percentage=40, letter_percentage=60, file="/usr/share/dict/american-english"):
    if len(ENGLISH_WORDS.keys()) == 0:
        load_language_file(file)
    counts = {'words':0, 'letters':0}
    words_in_msg = msg.split()
    for word in words_in_msg:
        if word.upper().strip(string.punctuation) in ENGLISH_WORDS:
            counts['words']+=1
    for char in list(msg):
        if char in string.ascii_letters:
            counts['letters']+=1
    return counts['words']/len(words_in_msg)*100 > word_percentage \
        and counts['letters']/len(msg)*100 > letter_percentage

Coin Marketplace

STEEM 0.17
TRX 0.16
JST 0.029
BTC 75912.93
ETH 2909.01
USDT 1.00
SBD 2.60